Paper 3-Measuring Homophily in Social Network Identification of Flow of Inspiring Influence Under New Vistas of Evolutionary Dynamics

(IJACSA) International Journal of Advanced Computer Science and Applications,
Special Issue on Extended Papers from Science and Information Conference 2013

14 | P a g e
www.ijacsa.thesai.org
Measuring Homophily in Social Network:
Identification of Flow of Inspiring Influence under
New Vistas of Evolutionary Dynamics
Hameed Al-Qaheri
Department of Quantitative Methods and
Information Systems
Kuwait Universit, Safat, Kuwait

Soumya Banerjee

Department of Computer Science
Birla Institute of Technology, Mesra,
India
AbstractInteraction with different person leads to different
kinds of ideas and sharing or some nourishing effects which
might influence others to believe or trust or even join some
association and subsequently become the member of that
community. This will facilitate to enjoy all kinds of social
privileges. These concepts of grouping similar objects can be
experienced as well as could be implemented on any Social
Networks. The concept of homophily could assist to design the
affiliation graph (of similar and close similar entities) of every
member of any social network thus identifying the most popular
community. In this paper we propose and discuss three tier data-
mining algorithms) of a social network and evolutionary
dynamics from graph properties perspective (embeddedness,
betweenness and graph occupancy). A novel contribution is made
in the proposal incorporating the principle of evolutionary
dynamics to investigate the graph properties. The work also has
been extended towards certain specific introspection about the
distribution of the impact, and incentives of evolutionary
algorithm for social network based events. The experiments
demonstrate the interplay between on-line strategies and social
network occupancy to maximize their individual profit levels.
KeywordsHomophily; Affiliation; Embeddedness;
Betweenness; Graph occupancy; Evolutionary dynamics
I. INTRODUCTION
Different properties of social network have demonstrated
potential interplay between events, participants and social
network itself. There are numbers of instances, where,
attribute of social network could drive the application area of
the network itself [21]. Hence, certain properties of social
network have an emerging impact [27] and homophily like
behavior is definitely one of them. Prior research
demonstrates impressive role of such behavior on the
application and analysis of social network [ 28] .
Homophily [1] [5] [6] [20], the tendency of individuals to
form association with individuals of similar socio-cultural
background, becomes the basic governing structural
component of any social network and it has been the focus of
many social network studies [2] [7]. Social network studies
reveal that social networks are homogeneous with regard to
many socio-demographic, behavioral and interpersonal
characteristics [3]. In any existing social network such as
Facebook or Twitter there lies some common functional
attributes such as posting photos, sending messages, likes,
dislikes, etc.. These kinds of activities lead to the concept of
affiliation towards a community [20]. From influence of social
propagation, Facebook and Twitter are dedicated to
disseminating the information and thus the concept of Twitter
follower graph and cascading of influence also reinforces the
hypothesis of different influence measurement model [4].
Considering the broader definition of the problem, this paper
finds a close similarity between graph theory and a social
network homophilic structure and explores the empirical
significance of influence propagation or a popular community
ranking and detection. The paper validates the existing graph
postulates with a proposed mining algorithm and simulation
applied a Facebook data set. Investigation yields certain
significant results with regards to popular community structure
and ranking based on different classical graph theory
properties like path traversal, Betweenness and Embeddedness
of social network nodes. After intial validation through graph
simulation, an initiative has been solicited with a self-
organizing and evolutionary principle, which could
dynamically trace the variants of social network. The role of
evolutionary dynamics [20] is also considerably significant as
it is defined as a study of the mathematical principles
according to which life has evolved and continues to evolve.
The evolution is also visible in the formation of social graph.
Reinforcement and validation of graph properties has been
demonstrated through the algorithmic strategies coined from
evolutionary dynamics [13] [14].The remaining part of the
paper is organized as follows: Section 2 elaborates the
statement of the problem with the parameters of graph theory
followed by existing methodologies, examples of social and
affiliation graph in section 2A and motivation of the analysis
has been discussed in section 2B. Section 3 describes
mathematical treatments responsible for proposed algorithm,
presented in section 4. Section 5 discusses the data set for
experiments and their implication on the graph properties of
social network. Section 5.1 introduced the role of evolutionary
dynamics to validate the simple graph properties for social
network instances, which may implicate in the mining of
graph related inferences. Finally section 6 gives conclusion
and mentions further scope of relevance research on the
paradigm.
II. STATEMENT OF THE PROBLEM
A community is formed in order to propagate or transfer or
share different knowledge across the network. And to disperse

15 | P a g e
the utility of ones community, it is required to apply some
techniques, such as voting or polling, on which knowledge
about the particular community has to be shared. When
forming community, certain basic parameters along with the
links among the members of the community are to be
considered. Parameters and links are as follows:
- Friend list
- Community links
- Graph occupancy period: Initial time and final time
- Paths and connectivity
- Path traversed
- Embeddedness
- Betweenness of nodes
- Affiliation
- Co-evaluation of social and affiliation network
A. Friend List
Friendship is developed in a social site based on some
common factor, like as, members belonging to same school,
working area, community and so on. Similar type of
characteristic people can be blocked into a common structure
and even a few from the block can also belong to some other
structure based upon their choice. [8] The homophily test of
friendship in a social site can be interpreted with the help of
the following example. Let there be a network where m
fraction of all individuals are male and f fraction of all
individuals are female. Considering a given edge in this
network, if we independently assign each node the gender
male with probability m and the gender female with
probability f then the both ends of the edge will be male
with probability m
2
and similarly both ends will be female
with probability f
2
. But if the first end of the edge is male
and the second end is female or vice versa then there exist
cross-gender edge. This condition will take place with a
probability 2mf. Thus the test for homophily according to
gender can be summarized as if the fraction of cross-gender
edges is less than 2mf then there is a presence of homophily
[8].
B. Community Links
Community can be created by a group of members by
selection and social influence method. The tendency of people
to form friendships with others who are like them are termed
as selection [8]. The selection criteria are mainly race or
ethnicity or similar characteristics. People may also modify
their behaviors to bring them more closely into alignment with
the behavior of their friends. This process is vividly described
as socialization and social influence [8].
The individual similar characteristics drive the formation
of links but social influence is a mechanism by which the
existing links in the network serve to share peoples
characteristics.
C. Graph occupancy period: Initial time and final time
It is the amounts of time spend on visiting a node. The
duration of remaining in a focus is calculated by checking the
difference in final time and initial time. This indicates the
graph occupancy period.
D. Paths and connectivity
According to the social scientists John Barnes defines
graph theory as Terminological jungle, in which any
newcomer may plant a tree.[8] A path is defined to be as a
sequence of nodes with the property that each consecutive pair
in the sequence is connected by an edge. The paths can also be
analyzed as not just the nodes but also the sequence of edges
linking these nodes. Connectivity can be described by saying a
graph is connected if for every pair of nodes, there is a path
between them. If a graph is not connected, then it is separated
into a set of connected pieces. Connected components of a
graph are a subset of the nodes such that the following two
properties hold [8].
1) Every node in the subset has a path to every other [8]
2) The subset is not part of some larger set with the
property that every node can reach every other [8].
E. Path Traversed
The path traversal is important in spread of important
information. It is required to examine whether something
flowing through a network has to travel just a few hops or
more. The Length of a path is the number of steps it
contains from beginning to end that is the number of edges in
the sequence that comprises it. The path traversal technique
used over here is the Breadth - First - Search. The method
of the traversal is one just need to keep discovering nodes
layer-by-layer, building each new layer from the nodes that are
connected to at least one node in the previous layer. Since it
searches the graph outward from a starting node, reaching the
closest node first, it is named as breadth-first-search [8].
F. Embeddedness
The number of common neighbors the two end points in a
network has referred to as embeddedness of an edge. This is
illustrated with the help of a schematic diagram in schema1.

Schema 1 An affiliation network Case
Here, embeddedness for two node A and node B has two
common neighbors node G and node H. Thus the concept of
embeddedness provides information that if two individuals are
G
H
B
D
C
A
E
F

16 | P a g e
connected by an embedded edge then this makes it easier for
them to develop a trust level and generate confidence for
transferring vital information or interacting with each other
[8].
G. Betweenness
Betweenness of a node is explained as the total amount of
flow that it carries when there exists a unit of flow between
each pair of nodes is divided up evenly over shortest path.
Nodes with high betweenness occupy critical roles in the
network structure. To compute betweenness efficiently we use
the notation of breadth-first-search. For a give graphical
structure the calculation of betweenness is done on the
perspective of time. For each given node the total flow from
that node to all other is distributed over the edges. This
technique is applied on every node in order to simply add up
the flow from all of them to get the betwenness on every edge
[8](shown in schema 2).

Schema 2: Local betweenness(the local betweenness of actor 1 is 2
H. Affiliation
Affiliation, a concept that is associated with homophily
graph, [8] [9] can be used to represent the participants, i.e. a
set of people, in a set of foci (representing some kind of
community). For example, node A, representing a person
could participate in focus X through an edge. These kinds of
graph are said to be affiliation network, since it represents the
affiliation of people (on left) with foci (on right). Affiliation
network is one of the examples of the bipartite graph.
Bipartite graph: A graph is said to be bipartite if its nodes
can be divided into two sets in such a way that every edge
connects a node in one set to a node in the other set [8].
Scheme 3 is an example showing nodes A and B representing
people participating in and Literature Club and Soft
Computing ) foci.
I. Co-evolution of social and affiliation networks
New friend links are formed and people become associated
with new foci over the period of time.

Schema 3 Affiliation through bipartite graph.
This kind of formation can lead to a kind of co-evaluation
which might indicate the selection choice of each individual
and there social influence. For example if two people belong
to a same focus then there is a probability that they become
friends and can influence each other with their community
they belong. According to the graph theoretic representation
nodes are used as both people and foci but the difference is
created by distinct type of edges. Firstly, an edge in a social
network, it connects two people and indicates friendship.
Secondly, an edge in an affiliation network, usually known as
social-affiliation network. This edge connects a person to a
focus and designates the operation of the person in the focus.
These two parameters can be resembled in the following
schema 3.

Schema 3 (a) When A, B, C are three different persons.

Schema 3 (b) A and B represent people but C denotes focus.
III. RECENT TRENDS AND MOTIVATION
While exploring the incentive generation, the basic
economics model of incentive distribution has become crucial
trend to be studied. Research already revealed the impact of
incentives on worker self-selection in a controlled and
restricted laboratory experiment. Subjects face the choice
between a fixed and a variable payment scheme. Considering
the status of the treatment, the variable payment is a piece rate,
a tournament, or a revenue-sharing scheme [22] [23].

B
Person

17 | P a g e
The extension of applying the graph properties of social
network is broadly inspired by Julia Poncela Casasnovass
research [24]. She addressed the study of the evolution of
cooperation on complex networks, using among the different
social dilemmas. Emphasizing mainly on the Prisoners
Dilemma game as a metaphor of the problem, her research
analysed possible outcomes of the dynamics, depending on the
underlying topology. Very recently, in 2013, it has been
pointed out that the topology not only highlights homophily or
other associated properties but also it leads toward role
discovery problem. role discovery problem [25] finds groups
of nodes that share similar topological structure in the graph.
But is it only topology or economies of incentive to quantify
the distribution of pay off under network? We investigated
from significant work on evolutionary games of Nowak et. al.
[26] and found that even payoff determines reproductive rate
and successful individuals have a higher payoff and produce
more offspring. Still it cannot be assured that the payoff also
could be signified by the carrying capacity of individual
participants in networked games. Finally, this work adopts
the emerging strategies of evolutionary game based incentive
distribution for any instance of social network under test.
IV. EXPLORING MATHEMATICAL TREATMENTS
The most influential community (focus) can be determined
by the frequency of the clicks made by the individual nodes.
However, visiting a focus and being a follower and then a
member of that focus are two different aspects. Visiting a node
could mean only collect information while being a member
means making the community well know. As such, to find the
most influential focus we need to find all possible paths from
the source to the destination and finding the shortest path by
computing the Betweenness values. Considering the graph one
node at a time, the distribution of the total flow over the edge
from that node to the other nodes could be computed. And
hence, the betweenness of the every node could be calculated
as follows:
Betweenness of every node = flows from different nodes
(1)
The shortest path traversed from the source to the
destination can be found by the Breadth-First Search
algorithm. Hence the number of shortest path to each node
should be the sum of the number of shortest path to all nodes
directly above it in the breadth-first search [8]. Valuation of
the members of the community or focus can be denoted by the
time spent on it by a particular node. Here lies the concept of
Graph Occupancy, which can be calculated by calculating the
time difference between Final time of leaving the focus and
Initial time of entering the focus. Mathematically, this can be
expressed as follows:
t
i
b f = /* frequency is non-linear relation with
betweenness. (2)
Where t={1,2,..m}
Since we concentrate on each node for calculating the
betweenness and thus finding their frequency of participants in
making a focus famous thus we have the following set of
Betweenness (B
i
) and Frequency (F
i
).This is expressed as:
{ }
n i
b b b B ,....... ,
2 1
= and { }
n i
f f f F .. ,......... ,
2 1
=
Equation 2, describes the frequency of acceptance of a
focus by different nodes. Based on a number of choices made
by each individual, we can select out the most beneficial or
famous (leader) communities, in terms of the population focus.
This entire logic can be expressed mathematically as:
( ) ( ) |
.
|
\
|
+ H + H =

=
C p F B p g
i
n
i
i
sin
1

(3)
where, p is the popularity of a node and C is a constant for
each delay unit process. In equation [3], the number of paths
to be followed to reach the destination is obtained by the
betweenness calculation denoted as
and frequency of the

particular node can also be obtained and denoted as F
i
. The
summation of these two values yields the minimum possible
path to be traversed to point the popularity of the community.
The maximum popularity of a community could be found by
the product of this result, in other words, by applying the
concept of Max-Min function, the maximum popularity of a
focus can be obtained for the minimum path travelled from the
source to the destination.
V. PROPOSED ALGORITHM -I
Finding the most influential community could be achieved
through the following algorithm which has distinct blocks to
evaluate the popularity and influence of node(s). The
algorithm contains certain unconventional graph properties
such as betweenness and embeddedness.
1. Begin
2. Initialization of links present in between the nodes and
focus
3. Initialization of variables
Betweeness=
, = {
}
Graph frequency= (
),
i
= {
}
Popularity P

4. Finding value
{t=1, 2, 3.n}
5. Calculating the betweenness:
Betweenness of every node= flows from different
nodes
6. Values form the betweenness leads to find the shortest path

7. Calculating the shortest path based on Breadth first search

8. Select a random node
which has visited a particular

community node at least once


18 | P a g e
9. while

do

Calculate the Graph frequency

if

then
F=

end if
Calculate g(p)=
+C)
10. end while
11. end begin
VI. IMPLEMENTATION AND ANALYSIS
The data set for validating the proposed model was
collected from Facebook community network. The reason to
choose Facebook was its close resemblance, in terms of nodes,
betweenness and edges, to the classical graph theory models.
A set of nine homophilic nodes, at a particular time, were
selected for initial validation. In these sets, each node is
connected to its influential nodes, which is denoted by the
links and their weights. There are communities that belong to
the nodes and each of these nodes tries to promote their own
communities.
To denote the most influential focus or community, we
need to find the shortest path (which also motivates the other
nodes to join the community and increase the popularity).
Embeddedness of the nodes are also considered to identify the
common neighbors among them. Table 1 shows the
parameters to be considered for the proposed algorithm. The
output value in the table was computed after implementing the
algorithm (using MATLAB version 7.10.0.499 (R2010 a)).
The detailed explanations of the parameters are as follows:
Friend List: 9 different Nodes represent 9 different friends
with their communities
Link Present: edges among the nodes with weights.
Embeddness: AB edge having common neighbors.
There may be present or might not be present. If not present,
then denoted by NIL. If present then node number is given.
Betweenness: involves reasoning about the set of all
shortest paths, between pairs of nodes.
Graph occupancy: amount of time spent on a node.
Leading to the further popularity of a community and
increasing the node linkage. The corresponding outputs are
given in the following Table 1.
TABLE I. LIST OF NODES WITH THEIR CONNECTIVITY, FLOW OF
INFORMATION AND FURTHER INCREASE IN POPULARITY
Parameters
Frien
d list
Links
present

Embedd
edness

Betweenness
Graph
occupancy

Node
1
(4,5,6) Node 4
(1,5) 0.2100

(1,6) 1.5300
(1,1) 0.0541
(1,2) 0.0462
(1,3) 0.0474
(1,4) 0.0148
(1,5) 0.0151
(1,6) 0.1478

Node
2

(3,5,6) Node 3

(2,3) 0.5100

(2,5) 0.3200
(2,2) 0.0591
(2,3) 0.0793
(2,4) 0.0131
(2,5) 0.0213
(2,6) 0.0106

Node
3
(4,5,6,7) Node 4 (3,4) 0.1500
(3,1) 0.0046
(3,3) 0.0266
(3,4) 0.0145
(3,5) 0.0135
(3,6) 0.0081

Node
4
(1,3,5,6,8) NIL

(4,1) 0.4500

(4,6) 0.7900
(4,1) 0.0435
(4,2) 0.0221
(4,3) 0.0446
(4,4) 0.0591
(4,5) 0.0065
(4,6) 0.1234

Node
5
(1,2,3,4) NIL

(5,3) 0.3200

(5,4) 0.3600
(5,1) 0.0111
(5,2) 0.0135
(5,3) 0.0827
(5,4) 0.0381
(5,5) 0.0375
(5,6) 0.0194

Node
6
(1,2,3,6) Node 4

(6,2) 0.4100

(6,3) 0.2900
(6,1) 0.0028
(6,2) 0.0377
(6,3) 0.0352
(6,4) 0.0030
(6,5) 0.0119
(6,6) 0.0591

Node
7
(3) NIL

(7,3) 0.4500
(7,2) 0.0191
(7,3) 0.0267
(7,4) 0.0046
(7,5) 0.0070

Node
8
(9) NIL

(8,9) 0.5500
(8,2) 0.0466
(8,3) 0.0652
(8,4) 0.0113
(8,5) 0.0171

Node
9
(4,8)

NIL

(9,4) 0.3000
(9,1) 0.0092
(9,3) 0.0531
(9,4) 0.0290
(9,5) 0.0271
(9,6) 0.0162


19 | P a g e

Fig. 1. Nodes with their links and weights
Figure 1 represents nine different influential nodes with
their links represented in the directed graph format. Each of
the edges has their weights marked on it. Here, the shortest
path has been considered from a source node say Node1 and
the destination node, Node 6.Time duration, that is the amount
of time spent by the nodes in their communities, can be an
invoking factor to join their community. Using equation (3),
Figure 2demonstrates that the most influencing or inspiring
node enhances the number of functionally active node. In this
figure, it is shown that the inspired nodes have added on more
of the connectivity with other different nodes.
To crawl Facebook, we implemented a distributed,
multithreaded crawler using Python with support for remote
method invocation (RMI) [11]. Facebook provides a feature to
show 10 randomly selected users from a given regional
network; we performed repeated queries to this service to
gather 50 user IDs to seed our breadth-first searches of
social links on each network
1
[11].
A. Role of Evolutionary Dynamics
Based on the intial simulation, it has been demonstrated
that there exist a strong cohesive directions with graph theory
and social network in the context of homophilic community
detection.

1
http://code.google.com/p/crawl-e/

Fig. 2. Enhancements in new connections with the existing invoking nodes
We need further investigation for the specific attributes in
the context of more homophily identification from the
perspective of either directed or undirected graph. The
extension of the algorithm will be significant to quantify the
application specific investigation of homophily. The
extension also could revalidate the correlation of graph
properties under social network.
Inspired by the phenomenal contribution in evolutionary
game theory by Nowak and his colleagues, several non linear
characteristics have been started adopting the concept of
evolutionary dynamics [12 ] [13 ]. It is evident that
evolutionary dynamics are defined by nonlinear differential
equations and therefore can be imported for revalidating
complex graph and network with the growth of a social graph
as mentioned in the first part of the algorithm. An evolutionary
dynamic assigns each population game F an ordinary
differential equation [14] ) x ( V x
F
= on the simplex x. One
simple and general way to define an evolutionary dynamic is
via a growth rate function:
n
R : g
n
R x (4)
Here, g represents the (absolute) growth rate of strategy i
and it will as a function of the current payoff toreward the
strategy. The previous algorithm only considered the shortest
path, breadth first search and graph frequency. As
Evolutionary Dynamics can also retrospect the growth aspects,
propagation and mutation of message under any state of graph,
hence therefore the conventional graph properties have been
revalidated using the extended algorithm. It should be
mentioned that proposed algorithm tries to incorporate
potential strength of evolutionary dynamics for simulating

20 | P a g e
the behavior and growth of social graph structure. Series of
more precise plots have been accomplished in post simulation
of extended algorithm (Figure 4-8). Persistence across the
network also indicates the out degree and also any average
number of distinct tags of groups and of tag assignments of
users having k
out
neighbours could be evaluated from social
network as shown. The users, who have more contacts in the
social network, tend also to be more active in terms of tags
and groups. Average number of distributive tags denoted as n
t.
group tags on the specific message as n
g
and subsequently n
w

represents the list of predefined choices. Correlations between
the activity of participants and their number of declared
friends and neighbours can be identified: here also the k
out
neighbours have been taken into consideration. The data has
been log-binned (Figure 3): by definition a bin of constant
logarithmic width signifies that the logarithm of the upper
edge of a bin (x
i
+1) is equal to the logarithm of the lower edge
of that bin (x
i
) plus the bin width (b) [19]. Here, the symbols
indicate the average, and the error bars with near optimal 25
and 75 percentiles for each bin. The algorithm deploys the
Python 2.7.3, which was released on April 2012. The present
implementation also allows the feature of Automatic
numbering of fields in the str.format() method, which have
been reflected in the post implementation stage of algorithm.
Algorithm II: Graph Pattern (G, P (t), )
1. Define the initial state of the graph , i.e. define G
i
=
i
o
/*
i
o is the
2
Kronecker symbol*/
2. Solve for Pi(t), which provides functions P(t) and i(t)
/*P(t) : Probability of sample occupancy time for t
m

Evaluate:
i
i
i
P
dt
dY
dt
dP
= =
(5)
/* The probability Yi(t) that at least one mutation has
occurred while the system was at state i before time t*/
3. Initialize the system with N classes of social network
instances at time t = 0, with its intialstart of connection,
k
out :
out-degree of graph, occupancy on graph, and
termination instance:
evaluate:
=
i
m i i
m i i
m i
) t ( P
) t ( P
) t (
(6)
/* where: probability ) t (
m i
that the social network is
at state i, P(t): sample occupancy, represents probable
mutation rate of messages across participants according
to the proportion of occupation*/
4. Sample the next mutation time according to the
cumulative probability P(t). This can be done via the

2
It is simply a function on two variables, i and j which are
integers, when for each social network instances cardinality of
the variables is large, then it could assist to approximate
inference based with a constrained, lower complexity,
adaptively sized sum for the target cardinal value [10 ].
inversion method, such that the next time t
m
= P 1(r),
where r is a uniform random variable between 0 and 1.
5. Add t
m
to the current time of occupancy of participants
and betweenness.
6. Choose the specific score and plot according to their
respective nore transition of the network and update the
state of the system as per Step 4.
7. Remove extinct and redundant classes from the list and
reduce the number N of classes accordingly.
8. Return to Step 1 until finished.

Fig. 3. Distribution of tagging with out degree (kout)

Fig. 4. Degree Correlation and distribution of Message tagging
In case of large scale network like Facebook, the most
conventionally investigated mixing pattern involves the degree
(number of neighbours) of nodes. This type of mixing
improvises the likelihood; leading users with a given number
of neighbours connect with users of similar degree. This
property is emphasized by computing multi-point degree
correlation functions. Complementary cumulative conditional
distributions as mentioned with group tagging, specific
message tagging and pre-defined choice tagging, compared
with the global cumulative distributions denoted by black lines
(Figure 4). Even among the subset of users with a given k
ou
t, a
strong disparity is still observed in the amount of activity and
also around a specific community. Subsequently, the

21 | P a g e
occupancy on a specific graph instance results in the following
plot: Log-log plot of the distribution of the contact durations
and of the cumulated duration of all the contacts two
individuals m and n have over a day (w
mn
). An interesting
inference could be drawn that out of 88% of the total contacts
sustained less than 1 minute on a specific tag or comments, but
more than 0.2% persisted more than 5 minutes against a
specific topic of interest. For the cumulated durations, 64% of
the total duration of contacts between two individuals during
one day last less than 2 minutes, but 9% last more than 10
minutes and 0.38% more than 1 hour. The small symbols
reciprocate to the actual distributions, and the large symbols to
the log-binned distributions [18] (Figure 5).

Fig. 5. Occupancy Graph for Social Network with Mutation Probability of
Messages
Degree distributions of evolutionary dynamics and the
empirical data: o represents the data points from the
simulated data and x represents the data points from the
empirical data new project The results of the degree
distributions of the extended algorithm with dynamic values
of mutation are shown in Figure 6. The upper figures are the
comparison of developer and project degree distributions in
linear coordinates.
The lower figures are the comparison of developer and
project degree distributions in log-log coordinates. The R
2
of
developer degree distribution from the simulated data in lower
figure is 0.959 and the R
2
of project degree distribution from
the simulated data in lower figure is 0.7657. Also the largest
project size of the simulated data is just 1500. We can further
lower this value by tuning the mutation parameter [17] and
P (t),and graph function of the extended algorithm. H log
represents homophily log evaluated from message exchanged
towards any specific and common interest [17].

Fig. 6. Evolutionary Dynamics plot with varying mutation and Probability of
occupancy

Fig. 7. Recursive simulation on test function towards Homophily from
Betweenness
In order to quantify the complex large cardinal network,
we incorporate classical test function like DeJong
3
(The
conventional De Jongs functions is the so-called sphere
function, this function is unimodal and convex by nature)[
16].Dimension is represented in the formulas by variable D, so
as can be observed that it becomes simple to calculate selected
functions for an arbitrary dimension for arbitrary number of
participants. Similarly, the result of homphily structure from
intial betweeness and other asscoited properties is safely
turned out from local mimima due to the size of its popluation
and avaiable best solution.The extended part of the algorithm
demonstrates the frquencey of interplay and incentive
distribution betwwen different communities played in social
network.

3

represents Dejong Function

22 | P a g e
Algorithm III Distribution in Population graph
The working strategy
Variables: N./*Homogeneous population of size*/
t /* Time Step t*/
r /* relative fitness */
/*temporal event could be Binomial time step
4
(equation 8)
of propagation of base event under social network , if
randomly chosen individual from SN then fixation of
probability of new mutant
1
represents probable mutation
rate of messages across participants according to the
proportion of occupation*/
N
1
1/r 1
1/r 1
=
(7)

/N 2i!
n
1) 2(i
4
1 i
2
i
N
1
t .... t t ][t
1/r 1
1/r 1
[ + + + +
= + + (8)
The distribution of trend on population graph of the new
attributes concerning the people refers category of homophily
structures (shown in red line) where the majority of people in
the population have lower levels of the attribute. Gradually, as
time elapses, the most influential message/ broadcast
configuration propagates across the population (indicated by
green line) but subsequently; evolution dynamics pushes the
higher level of the attribute with certain incentive. This is
natural recipient of incentives under social network for those
persons who are participating and interacting. In addition to,
there are drifts of incentive distribution under different
settings. Hence, eventually after a considerable period of
evolutionary dynamic, drift increases at the high end of the
attribute for homophilic structure and the distribution
approaches reverts back again to normal.

Fig. 8. Time versus messages

4
Refer to the Appendix for Proof

Fig. 9. Simulation of Pay off distribution
At this stage of simulation, it was not evident how the
appropriate interaction could entail higher range of incentive
among the participants. Therefore, a specific data set was
chosen to exhibit different interplay among the social
participants. The observations show that the frequency of
cooperators (indicated by blue), deviators (indicated byred)
and loners (represented as yellow) under smooth participation.
This is also measured as the multiplication factor according to
trust of discussion and interaction. Individuals are arranged on
random regular graphs where each node has eight neighbors
and they interact in randomly formed groups of size N = 5. For
small multiplication factors,individuals dominate. The reason
is simple: in this case even in a group of cooperators, the
payoffs do not exceed the incentive of individual. All of the
proposed components of three algorithms improvise dynamic
equilibrium considering the population of social network. The
range of multiplication with trust increases the pay off a very
small discrimination ofr.And above the threshold deviators
exist.. Only for much larger value of r~4.056 cooperators
reappear and co-exist with deviators. Since individuals are
absent, the dynamics again retain voluntary participation into
compulsory interaction. Finally, for r= 4.6 cooperators take
over and manage to displace deviators (see Figure 9 and table
1).

Table 1. Configuring design of Incentives on SN
The proposed design involves 4 stages and 4 groups
described below and summarized in the following Table:

23 | P a g e
- Group 1: This group comprises of more than 65
subjects. The default initial value of the participating
reward is 10 units and the final value could be
measured as .
- Group 2: This group represents almost 50 subjects
with initial incentive but in subsequent sessions the
experience differs with different fractions of incentives
depending on the frequency of interaction as shown in
Figure 8.
- Group n: The group enhances its subject line > 100, but
in addition to the normal incentive distribution, there
will be different treatments for incentive either there
should be donation or additional amount the
participants will push into it.
B. Comment on final Homophily Sturucure
The homphily evlalution from the social graph is
completely based on the increase of posts, tags and other
social network action artifacts. Keeping in mind about the
non-linear aspects and growth strategy of socail network,
evolutionary dynamics, comparatively better visualization of
density of attributes has become possible. Especially,
betweenness and graph occupancey and mutation could be
the one of those key factors. Recent research of Tom A.B.
Snijders and his colleagues [15] demonstrated the relation
between evolutionary dynamics and homophily of social
network in some common aspects of friendship,
recommendation, group study and selection etc. Figure 7
exhibits trend of precise visualization of homophily structure
as guided by the extended algorithm with mutation as prime
parameter. The following table 2 should be considered to
understand the relationship shown in Figure 8: the red circle
signifies higher cluster density and therefore providing the
remaining count of homophily over the population.
Population variables Mean Std. Dev. AutoCorr. Homophily
St.No.
Record Set tracking
From
snap.stanford.edu/data/

Male and Female 0.53 0.51 0.075 55 240000
Age
Density of Propagation
28.76 8.59
24.60 13.21
0.365 0-160
0.35 160-240

197000
230615

Network Variables

Out Degree
Local Betweeness
301.4 177.21
34.74 91.76
0.15 58- 90
0.017 170>
210043
228456

VII. CONCLUSION AND FURTHER SCOPE OF RESEARCH
Affiliation being an important factor of homophily graph
has a relevant role in promoting the focus through the nodes.
The idea is how a friend in the Facebook or any social network
can influence his friends to join the community he or she
belongs to. This paper investigates such possibilities by
exploring the homophily community in social network by
using graph property based algorithm and simulation. Further,
incorporation of evolutionary dynamics also contributed for
investigating homophily property with better approximation.
Subsequently, the growth of social network, temporal behavior
and trend of it, can also be investigated by augmenting the
existing evolutionary dynamics algorithm. On the other hand,
the ranking algorithm or conventional classifier model can
also be extended using initial attributes of embeddedness and
graph traversal with graph occupancy time. Soft computing
based (fuzzy and rough set) homophily identification from
graph properties could be an emerging research on
computational social network. Finally, in the extended
analysis part, certain significant observations are made in
terms of maximizing benefit or profit, based on their role at
specific instances under social network. Experiments have
incorporated certain public data sources to demonstrate mutual
interplay of social network participants, their specific
contribution towards the network and share of incentive if any.
REFERENCES
[1] Vojtek, P., &Bielikov, M. Homophily of neighborhood in Graph
Relational Classifier. Pages 721730 of: Geffert, V.,Karhumki, J.,
Bertoni, A., Preneel, B., Nvrat, P., &Bielikov, M. (eds), SOFSEM
2010: 36th Conf. on Current Trends in Theory and Practice of Computer
Science, SpindleruvMlyn, Czech Republic, 2010, Proc. LNCS, vol.
5901. Springer.
[2] Halil Bisgin, Nitin Agarwal and Xia oweiXu, Investigating Homophily
in Online Social Network, 2010,IEEE/WIC/ACM International
Conference on Web Intelligence and Intelligent Web Technology, pp.
533-536.
[3] M. McPherson, L Smith Lovin and J.M. Cook, Birds of a feather:
Homophily in Social Network, Annual Review of Sociology, 27(1): pp.
415-444, 2001.
[4] Eytan Bakshy, Winter A mason, Jake M.H .Ofman, Duncan J. watts,
Everyones an Influencer : Quantifying Influence on Twitter , published
in Proceedings WSDM 2011, Hongkong, February 2011.
[5] Reza Zafarani, William D Cole, and Huan Liu. Sentiment Propagation in
Social Networks A Case Study in LiveJournal. In Patricia Chai, Sun-Ki
And Salerno, John And Mabry, editor, SBP 2010 - Advances in
SocialComputing, Lecture Notes in Computer Science 6007, pp. 413
420. Springer Berlin, 2010.
[6] James H. Fowler, Jaime E. Settle, and Nicholas A. Christakis.Correlated
genotypes in friendship networks.PNAS, pp. 108:1993, 2011.
[7] Cosma Rohilla Shalizi and Andrew C. Thomas. Homophily and
Contagion Are Generically Confounded in Observational Social
Network Studies. Sociological Methods and Research
(arXiv:1004.4704v3), 2011.

24 | P a g e
[8] Networks, Crowds, and Markets: Reasoning about a Highly Connected
World. By David Easley and Jon Kleinberg.Cambridge University Press,
2010.
[9] Benjamin Goluby Matthew O. Jackson, How Homophily Affects the
Speed of Learning and Best-Response Dynamics, Quarterly Journal of
Economics Vol. 127, Iss. 3, pp 1287--1338, 2012.
[10] Chris Pal, Charles Sutton, Andrew McCallum. Constrained Kronecker
Deltas for Fast Approximate Inference and Estimation.Submitted to
UAI, 2005.
[11] Bryce Boe and Christo Wilson.User Interactions in Social Networks and
their Implications,
accessableat http://cs.ucsb.edu/~bboe/public/pubs/interaction-
eurosys09.pdfcrawl-e: Highly distributed web crawling framework
written in python. Google Code, 2008.
[12] Chatterjee K, D Zufferey, MA Nowak (2012). Evolutionary game
dynamics in populations with different learners.JtheorBiol 301: 161-173.
http://dx.doi.org/10.1016/j.jtbi.2012.02.021
[13] Allen B, MA Nowak (2012). Evolutionary shift dynamics on a
cycle.JtheorBiol 311: 28-39.
http://dx.doi.org/10.1016/j.jtbi.2012.07.006.
[14] Survival of dominated strategies under evolutionary dynamics, Josef
Hofbauer and William H. Sandholm,
Theoretical Economics 6 (2011), pp. 341377.
[15] Alessandro Lomi, Tom A.B. Snijders, Christian E.G. Steglich, and
Vanina Jasmine Torl (2011). Why Are Some More Peer Than Others?
Evidence from a Longitudinal Study of Social Networks and Individual
Academic Performance. Social Science Research, 40 (2011), 1506-1520.
[16] X.-S. Yang, Test problems in optimization, in: Engineering
Optimization: An Introduction with Metaheuristic Applications (EdsXin-
She Yang), John Wiley & Sons, 2010.
[17] Y. Gao,G. Madey, V. Freeh, Modeling and Simulation of a Complex
Social System: A Case Study, accessible at
http://www3.nd.edu/~oss/Papers/SimulationSym37_gao.pdf
[18] J. Stehl, and Et al, High-Resolution Measurements of Face-to-Face
Contact Patterns in a Primary School.Published: Aug 16, 2011,DOI:
10.1371/journal.pone.0023176
[19] E. P. White, B. J. Enquist, and J. L. Green. Ecological On estimating the
exponents of power-law frequency distributions. Archives E089-
052-A1. 2008. Ecology 89:905-912.accessable at
http://www.esapubs.org/archive/ecol/E089/052/appendix-A.pdf
[20] Al-Qaheri, H, Banerjee , S., Ghosh, G. Evaluating the power of
homophily and graph properties in Social Network: Measuring the flow
of inspiring influence using evolutionary dynamics, Science and
Information Conference (SAI), London, 2013 , IEEE XPlore, pp. 294
303.
[21] Himabindu Lakkaraju, Julian McAuley, Jure Leskovec Whats in a
name? Understanding the Interplay between Titles, Content, and
Communities in Social Media, Association for the Advancement of
Artificial Intelligence ( 2013
[22] Dohmen, T., and Falk, A., Performance Pay and Multi-Dimensional
Sorting: Productivity, Preferences and Gender, American Economic
Review, 101(2), pp. 556-590, 2011.
[23] DellaVigna, S., J. A. List, and Malmendier U., Testing for Altruism
and Social Pressure in Charitable, Giving, Quarterly Journal of
Economics, 127(1), pp.1-56. 2012.
[24] J. Poncela Casasnovas, Evolutionary Games in Complex Topologies,
Springer Theses DOI: 10.1007/978-3-642-30117-9, Springer-Verlag
Berlin Heidelberg 2012.
[25] Sean Gilpin, Tina Eliassi-Rad and Ian Davidson , Guided Learning for
Role Discovery (GLRD): Framework, Algorithms, and Applications
KDD13, August 1114, 2013, Chicago, Illinois, USA.
[26] Sebastian Novak, K. Chatterjee and M.Nowak, Density games Journal
of Theoretical Biology 334, pp.2634, 2013.
[27] Himabindu Lakkaraju, Indrajit Bhattacharya, Chiranjib Bhattacharyya,
Dynamic Multi-Relational Chinese Restaurant Process for Analyzing
Influences on Users in Social Media .IEEE International Conference on
Data Mining (ICDM), 2012.
[28] Debanjan Mahata and Nitin Agarwal. Grouping the Similar among the
Disconnected Bloggers, Social Network Analysis and Social Media
Mining: Emerging Research. Guandong Xu and Lin Li (Eds.). IGI
Global, 2012.
APPENDIX: PROOF OF STATEMENT IN EQUATION 8:
The proposed model is to trace the time distribution mode
in between different intermediate events. Considering the
normal binomial distribution function, which specifies the
number of times (x) that an event occurs in n independent
trials where p is the probability of the event occurring in a
single trial. Why we considered this trial? As, the exact
probability distribution for any number of discrete trials may
represent the number of mutant messages, therefore it
becomes obvious that for any discrete time t, say for any
temporal event the value of number of events could be large.
Hence function becomes continuous. Thus relative fitness r
also changes respective to bi-nominal distribution of mutant
messages.
Hence:
/N 2i!
n
1) 2(i
4
1 i
2
i
N
1
t .... t t ][t
1/r 1
1/r 1
[ + + + +
= + +
\
is a normal representation of r with bi-nomial time t

Paper 3-Measuring Homophily in Social Network Identification of Flow of Inspiring Influence Under New Vistas of Evolutionary Dynamics

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Paper 3-Measuring Homophily in Social Network Identification of Flow of Inspiring Influence Under New Vistas of Evolutionary Dynamics

Uploaded by

Copyright:

Available Formats

(IJACSA) International Journal of Advanced Computer Science and Applications,

and frequency of the

which has visited a particular

You might also like